armansportal.com · personal playground

Arman's
playground▮

Not a brand, not a blog, not a funnel. This is where my experiments live: trading research with receipts, and benchmarks I ran because I wanted the real answer — not the marketing one.

select a world save data is real · nothing here is advice
slot 01futures desk

Trading

A three-year engineering journal pointed at MES futures: one verified method, an 11,112-variant research database, prop-firm rule batteries, raw replay tape, and simulation labs.

 one method · 7 firms
last save 2026-08 → slot 02agent bench

Agent Benchmark

Five coding agents, twelve tasks, rubric-graded with the receipts shown — because "it feels smarter" is not a metric.

 5 agents · 12 tasks
last save 2026-08 → slot 03jarvis trials

Hermes Benchmark

Local models auditioning for the Jarvis seat, on my own hardware. Most failed. The failures are published too.

 local llms · full artifacts
last save 2026-05 →